The Lesson from SARS - APAMI

1

Containing Acute Disease Outbreak

Digital Ring Fence: A novel GIS-based approach to contain

Emerging Infectious Diseases (EIDs)

K C LunImmediate Past President, International Medical Informatics Association (IMIA)and Professor,School of Biological Sciences,Nanyang Technological University, Singapore

presented by

27-29 October 2006, Taipei

APAMI 2006


The Lesson from SARS

First Quarter, 2003

SARS coronaviruses

2


Source: Ostroff, National Center for Infectious Diseases, CDC

“Even if we rapidly eliminate SARS, we remain at risk of future viral mutations and should expect more dangerous new viruses to emerge over the next ten years”.

- Patrick Dixon, futurist and chairman of Global Change Ltd, 7 May 2003

3


• Will H5N1 be the next scourge after SARS?

• Already spreading beyond Asia to Europe

• Spread possibly by infected migratory birds

• World governments anticipating (and preparing) for a global pandemic

4


Source: Ostroff, National Center for Infectious Diseases, CDC


SARS Control• Identify the sick people• Treat the sick people

without infecting others• Keep contacts of sick

people at home for 10-14 days i.e. Ring-fence (corral) the spreader and potential spreaders

Nurses under quarantine at Hoping Hospital, Taipei

5


Ring-fence strategy in disease epidemiology

• Denotes an activity to contain the spread of infection by – isolating contacts (exposure)– monitoring them for the

disease (infection)

Physical ring-fencing gave rise to idea of Digital (electronic) Ring-Fencing Technology (DRiFT)

• For SARS, Singapore Ministry of Health enforced strict quarantine of exposed through rigorous contact-tracing

Source: Carlos Castillo-Chavez, Center for Nonlinear Studies, Los Alamos National Laboratory, Cornell University

Classical Epidemiological Modeling

6


Modeling SARS transmission• Traditional “compartmental” modeling (SEIR) in

epidemiology assumes population groups are fully-mixed i.e. every individual has equal chance of getting the disease and spreading it to one another

• For a disease like SARS, the basic reproductive number (Ro) has not produced large-scale epidemics as predicted

• Compartmental modeling breaks down for diseases transmitted through close contact e.g. SARS


Contact Network Modeling• Discrepancy between estimates of Ro

and observed epidemiology possibly due to early and effective intervention

• Need to focus on “individual-based modeling” instead of “compartmental modeling”

• Recent work focuses on “contact network modeling” to characterize every interpersonal contact that can potentially lead to disease transmission in the community

7

Contact Networks to model spread

vertexedge

degree

.!

)1(k

eppkN

pk

kNkk

λλ−− ≈−⎟⎟⎠

⎞⎜⎜⎝

⎛=

Probability of 1 point having k degrees (no. of lines connected to it) =

where pN≈λ


DRiFT: Digital Ring Fence Technology for EID containment using GIS platform

Map occurrences of cases (vertices of contact network)Encircle each vertex with digital ring fenceRadius of digital ring fence proportionate to estimated risk of infection of individuals within the ring fenceGIS platform allows containment efforts to be directed at places of population concentrations (markets, shopping malls, entertainment centres) within geographical area marked out by digital ring fence Strategy: Make disease containment one-step

ahead of its spread

8


GIS mapping of outbreaks

Map location of vertices


Multi-layering of GIS maps

9


Multi-layering of GIS maps


DRiFT Focus: Contact network to model disease containment

• Given the vertex, what is the risk of someone in the contact network getting the disease within x distance from the vertex?

Model focuses on containment, not spread of the disease

R2

R1

• Distance from vertex and its concomitant risk provide basis for the digital ring fence construction Risk 2 (R2) < Risk1 (R1)

10


Digital Ring Fence: Model

!)]()([})]()({[ ))()((

nxyenxNyNP

nxy λλλλ −

==− −−

∫=x

dzzgx0

)()(λ

If the probability

where

then N(y)-N(x) has a Non-Homogeneous Poisson distribution with parameter )()( xy λλ −


Digital Ring Fence: ModelThe overall risk of infection within the defined area is given by:

.)(*))()((),(

xyPSTxNyNEyxr

−−

=

where PS(y-x) = the population size within the defined area and T = infectivity of the infected person

11


Digital Ring Fence: Model

Depending on the risk of infection calculated, will adjust the radius of the e-ring fence to cope with the containment of the disease spread

xy

z


Simulation Studies• Hypothetical data of

individuals coming into contact with a probable case over last 3 days are electronically captured

• Stored in secured databaseData 1 records the information regarding the distance between the probable case and his contacts on day 1

Data1 = {0.00126018392283 0.00893845993502 0.01293193288195 0.01721042081285 0.11229577204808 0.21482253481043 0.31507974052212 0.37614151271792 0.49735498028620 0.70294088286048 0.97109194057815 2.16890715160328 2.72096888046909 3.52297913844732 4.66629786044356 4.83623313936767}(kilometers)

12

• From data, piecewise-linear nonparametric estimator of the cumulative intensity function is derived

Simulation Studies

⎥⎥⎦

⎤

⎢⎢⎣

⎡ Λ+Λ

Λ−Λ

kxzx

kxzx )(ˆ)(ˆ,)(

ˆ)(ˆ 2/2/ αα

with confidence intervals given by:

• Estimator allows us to estimate average number of individuals coming into contact with probable case per day within the area of the digital ring fence

Estimator with 95% CI after smoothing

13


Computing Risk of Infection for DRiFT

• E(N(y)-N(x)) = estimated number of contacts with probable case within area of concentric rings with radii y and x

• T = transmissibility of infectious disease• PS(y-x) = population size within area of

concentric rings with radii y and x

.)(*))()((),(

xyPSTxNyNEyxr

−−

=

Advantage of the GIS – allows estimation of population size within the defined area


14

The Real Test• The digital ring fence concept is to guard

against the spread of infection• When an outbreak occurs, places within

the buffer zone on the GIS with high population concentrations can be put on high alert to keep the outbreak in check

• Real test comes when an outbreak does occur, especially if the mechanism of transmission goes beyond person-to-person contact


Summary: Handling EIDoutbreaks

Strategy:Containment not prediction of the course of outbreak

15


Containment Strategy:Construct digital ring fence and focus contain-ment efforts on places with high infection risks within the digital ring fence



Goal:Make disease containment one step ahead of its spread


16


H5N1 OutbreaksDigital ring fence concept can also be applied to delineate culling zone for poultry



17

Acknowledgement

Research team members:K C Lun (Principal Investigator)Antony Prince (Research Associate)Chen Xin (Research Associate)

Project Grant #M47080014 from Biomedical Research Council, Agency for Science, Technology and Research (A*STAR), Singapore


E-mail:[email protected]:[email protected]

Thank You !Thank You !谢谢！

Documents

The Lesson from SARS - APAMI